translation quality 0.00253222
system quality 0.002024683
score matrix 0.001965499
reliable human 0.0018636039999999998
human ratings 0.0018407279999999998
human assessment 0.0018054879999999998
human judgments 0.001784229
translation hypotheses 0.0017829200000000001
data set 0.001773382
human judges 0.0017547419999999999
human evaluations 0.0017523979999999998
score scale 0.001735795
human evalua 0.0017177079999999999
human judg 0.0017076689999999998
human assessments 0.001705449
human grades 0.00169965
single score 0.001652896
original score 0.001639517
single scores 0.001604506
multiple scores 0.001549254
full score 0.0015247070000000001
true scores 0.001524438
data techniques 0.001521329
only scores 0.001501774
actual scores 0.001497407
translation 0.00149328
entire score 0.0014918520000000001
simulated data 0.001490707
random matrix 0.001488535
medium score 0.0014841860000000002
hter score 0.0014841860000000002
covered score 0.0014841860000000002
eraged score 0.0014841860000000002
absolute scores 0.001478007
known scores 0.001465619
ing data 0.00145956
human 0.00143335
missing data 0.001427245
quality threshold 0.001424248
absolute error 0.001385197
data increases 0.001382106
ordinal data 0.001381438
different ways 0.001374689
recovery quality 0.0013436519999999999
nlp system 0.00133695
different strategies 0.001336318
different judges 0.001319523
quality estima 0.001315234
mean method 0.0012999930000000002
different order 0.001292194
different kind 0.001289201
different evalua 0.001282489
sentence level 0.001218638
score 0.00121782
automatic translations 0.001216131
experimental results 0.0011917999999999998
scores 0.00116943
original matrix 0.001169376
first corpus 0.001164094
matrix rank 0.001119036
matrix completion 0.001108764
hypothesis matrix 0.001097495
random noise 0.001096271
error 0.00107662
matrix com 0.001067257
random samples 0.00106137
strategies models 0.001056206
statistical measures 0.001055102
full matrix 0.001054566
quality 0.00103894
recovered matrix 0.001031211
evaluation 0.00101799
dicted matrix 0.001015326
inal matrix 0.001014578
pleted matrix 0.001014578
matrix repre 0.001014578
matrix comple 0.001014578
noise level 0.001006481
several experiments 0.001000939
alternative approach 9.97491E-4
mean baseline 9.937540000000001E-4
second corpus 9.91215E-4
other corpora 9.876239999999999E-4
system 9.85743E-4
several works 9.80533E-4
several judgments 9.75783E-4
prediction methods 9.646559999999999E-4
tify errors 9.56744E-4
other judges 9.52362E-4
negative results 9.51777E-4
rand method 9.473789999999999E-4
tides corpus 9.46162E-4
recommender systems 9.4403E-4
fluency corpus 9.43978E-4
large impact 9.42805E-4
simple mean 9.41024E-4
large annotation 9.40195E-4
mean technique 9.394570000000001E-4
pred method 9.35784E-4
prediction performance 9.27401E-4
